Package to handle model training for dpm tasks
Project description
Welcome to lightsaber
Overview
lightsaber
is designed ground up to provide a simple, composible, and unified
model training framework. It has been designed based on state-of-the-art open source
tools and extended to support the common use cases for disease progression modeling (DPM).
lightsaber
provides four main components:
- Data ingestion modules
- Model Trainers
- DPM problem specific model evaluation
- Model tracking and support for post-hoc model evaluation.
Each of these components are designed such that a user should be able to pick some
or all of the modules and embed these seamlessly with their current workflow.
Futhermore, when used in the recommended manner, lightsaber
provides a batteries included
approach allowing the modeler to focus only on developing the logic of their model and
letting lightsaber
handle the rest.
Currently, we support the following DPM use cases:
- classification: one or multi-class
Also, we support and extend the following frameworks:
scikit-learn
compliant models: for classical modelspytorch
compliant models: for general purpose models, including deep learning models.
To summarize, it is thus an opinionated
take on how DPM should be conducted providing with a
unified core to abstract and standardize out the engineering, evaluation, model training, and model tracking
to support: (a) reproducible research, (b) accelarate model development, and (c) standardize model deployment.
Installation Instructions
Lightsaber
is installable as a python package.
It can be installed using conda
as:
conda install -c conda-forge dpm360-lightsaber
or from pypi
as:
pip install dpm360-lightsaber
It can also be installed from source using pip as follows:
- barebones install of
Lightsaber
:pip install .
- with doc support:
pip install .[doc]
- with time-to-event modeling (T2E) support:
pip install .[t2e]
- full install with all components:
pip install .[full]
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for dpm360_lightsaber-0.3.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 18856ff57b2b38a2435ea1ba46f7e93ddfc6eaaf1772ca9dbbb2ea9364a125d4 |
|
MD5 | ac15db5f222fa63b382ca7a453d021f1 |
|
BLAKE2b-256 | 6aada9ffb2c50e29f2301cfc6b2226db0c69ea896c3405393d784a9ba273822e |